English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recently, Bird's-Eye-View (BEV) representation has gained increasingattention in multi-view 3D object detection, which has demonstrated promisingapplications in autonomous driving. Although multi-view camera systems can bedeployed at low cost, the lack of depth information makes current approachesadopt large models for good performance. Therefore, it is essential to improvethe efficiency of BEV 3D object detection. Knowledge Distillation (KD) is oneof the most practical techniques to train efficient yet accurate models.However, BEV KD is still under-explored to the best of our knowledge. Differentfrom image classification tasks, BEV 3D object detection approaches are morecomplicated and consist of several components. In this paper, we propose aunified framework named BEV-LGKD to transfer the knowledge in theteacher-student manner. However, directly applying the teacher-student paradigmto BEV features fails to achieve satisfying results due to heavy backgroundinformation in RGB cameras. To solve this problem, we propose to leverage thelocalization advantage of LiDAR points. Specifically, we transform the LiDARpoints to BEV space and generate the foreground mask and view-dependent maskfor the teacher-student paradigm. It is to be noted that our method only usesLiDAR points to guide the KD between RGB models. As the quality of depthestimation is crucial for BEV perception, we further introduce depthdistillation to our framework. Our unified framework is simple yet effectiveand achieves a significant performance boost. Code will be released.

BEV-LGKD: A Unified LiDAR-Guided Knowledge Distillation Framework for BEV 3D Object Detection